Fast high dimensional fixed effect estimation following syntax of the fixest R package. Supports OLS, IV and Poisson regression and a range of inference procedures. Additionally, experimentally supports (some of) the regression based new Difference-in-Differences Estimators (Did2s).
Project description
PyFixest
PyFixest
is a Python clone of the excellent fixest package. The package aims to mimic fixest
syntax and functionality as closely as Python allows. For a quick introduction, see the tutorial.
Functionality
At the moment, PyFixest
supports
- OLS and IV Regression
- Poisson Regression
- Multiple Estimation Syntax
- Several Robust and Cluster Robust Variance-Covariance Types
- Wild Cluster Bootstrap Inference (via wildboottest)
- Difference-in-Difference Estimators:
- Gardner's two-stage ("
Did2s
") estimator is available via thepyfixest.experimental.did
module
- Gardner's two-stage ("
Installation
You can install the release version from PyPi
by running pip install pyfixest
or the development version from github.
Benchmarks
All benchmarks follow the fixest benchmarks. All non-pyfixest timings are taken from the fixest
benchmarks.
News
PyFixest
0.10.8 adds experimental support for Gardner's two stage "DID2s" estimator:
import pandas as pd
import numpy as np
from pyfixest.experimental.did import did2s
from pyfixest.estimation import feols
from pyfixest.visualize import iplot
# download csv from this repo
df_het = pd.read_csv("https://raw.githubusercontent.com/s3alfisc/pyfixest/master/pyfixest/experimental/data/df_het.csv")
fit = did2s(
df_het,
yname = "dep_var",
first_stage = "~ 0 | state + year",
second_stage = "~i(rel_year)",
treatment = "treat",
cluster = "state",
i_ref1 = [-1.0, np.inf],
)
fit_twfe = feols(
"dep_var ~ i(rel_year) | state + year",
df_het,
i_ref1 = [-1.0, np.inf]
)
iplot([fit, fit_twfe], coord_flip=False, figsize = (900, 400), title = "TWFE vs DID2S")
Quickstart
from pyfixest.estimation import feols
from pyfixest.utils import get_data
data = get_data()
# OLS Estimation
fit = feols("Y~X1 | csw0(f1, f2)", data = data, vcov = {'CRV1':'group_id'})
fit.summary()
# ###
#
# Model: OLS
# Dep. var.: Y
# Inference: CRV1
# Observations: 998
#
# | Coefficient | Estimate | Std. Error | t value | Pr(>|t|) | 2.5 % | 97.5 % |
# |:--------------|-----------:|-------------:|----------:|-----------:|--------:|---------:|
# | Intercept | 2.206 | 0.078 | 28.304 | 0.000 | 2.043 | 2.370 |
# | X1 | 0.358 | 0.051 | 6.962 | 0.000 | 0.250 | 0.466 |
# ---
# RMSE: 1.765 Adj. R2: 0.024 Adj. R2 Within: 0.024
# ###
#
# Model: OLS
# Dep. var.: Y
# Fixed effects: f1
# Inference: CRV1
# Observations: 997
#
# | Coefficient | Estimate | Std. Error | t value | Pr(>|t|) | 2.5 % | 97.5 % |
# |:--------------|-----------:|-------------:|----------:|-----------:|--------:|---------:|
# | X1 | 0.411 | 0.040 | 10.188 | 0.000 | 0.326 | 0.495 |
# ---
# RMSE: 1.421 Adj. R2: 0.048 Adj. R2 Within: 0.048
# ###
#
# Model: OLS
# Dep. var.: Y
# Fixed effects: f1+f2
# Inference: CRV1
# Observations: 997
#
# | Coefficient | Estimate | Std. Error | t value | Pr(>|t|) | 2.5 % | 97.5 % |
# |:--------------|-----------:|-------------:|----------:|-----------:|--------:|---------:|
# | X1 | 0.431 | 0.035 | 12.319 | 0.000 | 0.358 | 0.505 |
# ---
# RMSE: 1.2 Adj. R2: 0.07 Adj. R2 Within: 0.07
Standard Errors can be adjusted after estimation, "on-the-fly":
fit1 = fit.fetch_model(0)
fit1.vcov("hetero").tidy()
# Model: Y~X1
# ###
#
# Model: OLS
# Dep. var.: Y
# Inference: hetero
# Observations: 998
#
# | Coefficient | Estimate | Std. Error | t value | Pr(>|t|) | 2.5 % | 97.5 % |
# |:--------------|-----------:|-------------:|----------:|-----------:|--------:|---------:|
# | Intercept | 2.206 | 0.088 | 25.180 | 0.000 | 2.034 | 2.378 |
# | X1 | 0.358 | 0.068 | 5.254 | 0.000 | 0.224 | 0.491 |
# ---
# RMSE: 1.765 Adj. R2: 0.024 Adj. R2 Within: 0.024
Last, PyFixest
also supports IV estimation via three part formula syntax:
fit_iv = feols("Y ~ 1 | f1 | X1 ~ Z1", data = data)
fit_iv.summary()
# ###
#
# Model: IV
# Dep. var.: Y
# Fixed effects: f1
# Inference: CRV1
# Observations: 997
#
# | Coefficient | Estimate | Std. Error | t value | Pr(>|t|) | 2.5 % | 97.5 % |
# |:--------------|-----------:|-------------:|----------:|-----------:|--------:|---------:|
# | X1 | 0.479 | 0.096 | 4.979 | 0.000 | 0.282 | 0.676 |
# ---
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for pyfixest-0.11.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 7b96b0f0706cc8a1fb688fb608179f06b86427b25f77972695ada756d98b9ac5 |
|
MD5 | 83b127c6b8b58a0d645c6a4bf52496ea |
|
BLAKE2b-256 | 3296cd23883949e825e2c5c9a0e5852ae6c440144928bacaf8558aab87e2d6a9 |